Download Tutorial Introduction

UAM CorpusTool
Version 3.0
Tutorial Introduction
(June, 2013)
Mick O’Donnell
[email protected]
1
About this Document
This document provides a tutorial introduction to UAM CorpusTool 3.0
(henceforth: UAMCT3). For more detailed information about the options in
each screen and menu of UAMCT3, please see the UAMCT3 User Manual.
About UAM CorpusTool 3.0
UAM CorpusTool is a set of tools for the linguistic annotation of text. Core
concepts include:
 The user defines a project, which is: a set of files, and a set of analyses
which are applied to each of these files.
 All the files of a project are stored in a single folder: the original texts (the
‘corpus’), the annotations on this text and the coding schemes (the tags
applied to the texts).
 Each ‘analysis’ can be seen as a ‘layer’ of annotation. CorpusTool
currently allows two types of annotation:
1. Document Coding: where the text as a whole is assigned features.
For instance, these features could represent the register of the
document (field, tenor, mode), or text-type.
2. Segment Coding: The user can select segments within a file, and
assign features to each of these segments. Segments are specified by
dragging the mouse over a span of text, and the user is then prompted
to specify the features of this segment.
 Annotation can be ‘manual’ (the user swipes text and chooses categories
for it) or ‘automatic’ (the program does the annotation for you). Sometimes
annotation is mixed, for instance, you can have the program recognise
clause or noun-phase segments, but it is up to the you to code them.:
CorpusTool is available from:
http://www.wagsoft.com/CorpusTool/
See that site for instructions on how to install CorpusTool on your machine.
2
Tutorial 1:
Starting a new project
1 Launch UAM CorpusTool
Once UAM CorpusTool is installed on your machine, you can begin working with
it. The first thing to do is to create a new “project”:
Windows:
 When installing CorpusTool, you had the option to place an icon on the
desktop. Click on this icon to launch CorpusTool.
 Alternatively, there should be a UAM CorpusTool icon in the Programs
menu in the Start menu on Windows Toolbar. Select this to launch
CorpusTool.
Macintosh:

The installation of CorpusTool placed the application in your Applications
folder. Double-click on the application to launch it.

You might find it useful to place the application in the Dock for easy
access.
If you have already created a project, you can open it simply by double-clicking
the .cp3 file in the Project folder. This file has an icon as below:
MacOSX:
Windows:
The Opening Window
A window should appear as in Figure 1.1. This window provides, amongst other
information, the version number you are using (useful if you need to
communicate bugs).
3
Figure 1.1: The Opening Window
The Window offers several options,
 Start New Project: create a new project from scratch.
 Open Project, to continue with a project you have already started, you will
be prompted to select one.
 Import Project from UAMCT 2: If you have a project from UAMCT 2, you
can use the “Import Project from UAMCT 2” button to make a copy of your
project in the UAMCT 3 format.
 Open SomeProjectName: If you have opened a project previously on this
machine, there will also be a button to open the last project opened.
2 Click on the “Start New Project” button.
After clicking this button, a “Create Project Wizard” will appear, which will lead
you through the steps needed to create your project:
1. Providing a name for a new project
2. Specify the folder where your new project’s folder is to be stored. For
instance, choose the Desktop folder on your machine.
When you click the “Finalise” button, CorpusTool will create your project, which is
a folder containing all the details related to your project, including the corpus, and
the annotation files. It also contains an icon which can be used to launch your
project directly (the .ct3 file).
Once you have finished with the Create Project Wizard, the CorpusTool Main
Window will open, showing the File pane. See Figure 1.2. This pane is where you
add or remove files to your project, or open a file for annotation.
4
Figure 1.2: The File Management pane
The buttons at the top of the pane allow you to switch between the different
panes of CorpusTool: Files (Tutorial 2), Layers (Tutorial 3) Search (Tutorial 5),
Autocode (Tutorial 6), Statistics (Tutorial 7), Explore (Tutorial 8), Options and
Help.
We will assume for now that the “File” pane is selected. The name of your project
is shown in the title bar of the Project window. In the space below is a box
showing all the files in the project (initially empty), and for each file, one button for
each of the possible analyses of that file.
This ends the first tutorial. The next tutorial will show how to add content to your
project.
5
Tutorial 2:
Adding text files to your project
The next step is to add some files to the project.
1 Save Documents as plain text
UAMCT 3 deals only with plain text files. If your files are in MS Word format or
PDF, you need to save them as plain text.
If you are on Windows, and your texts are in languages with non-western
characters (e.g., Cyrillic, Chinese, Korean, etc.), then it is better to open your
.docx document with WordPad, and use the “Save as…” option there, as it can
save as a Unicode file.
2 Click on “Extend Corpus”
Click on the Extend Corpus button in UAMCT. A window will appear to guide you
through the process of adding files. You are given a choice between:
 Add a single file: You will be asked to select a file to add to the corpus.
Additionally, you will be asked to specify a “subcorpus” for the text file.
Texts in UAMCT are stored within subcorpora (folders within the Corpus
folder). For instance, you might have one subcorpus for native texts, and
another for learner texts.
 Add a folder of files: You will be asked to select a folder to add to the
corpus. This folder could be:
o A folder of plain text files: the folder will be added as a “subcorpus”
of the project.
o A folder of folders of plain text files: each folder will be added as a
“subcorpus” of the project.
 Paste from the Clipboard: you will be given a space in which to
copy/paste text into. This is a useful way to take texts from the internet into
UAMCT.
In the first two cases, the files you select will be copied from where they are
into the Corpus folder of your project. The originals are left untouched.
For this tutorial, lets use the “Paste From Clipboard” option. Copy the following
paragraphs of text and follow the instructions below:
Obama is like Apple, Google and Facebook: a once hip brand tainted by Prism
Among the guests at the fabled Bilderberg meeting, held this weekend just
outside London, are the top brass of Google, Amazon and Microsoft. How
appropriate they should be there, alongside luminaries of the US political and
military establishment. For this was the week that seemed to confirm all the old
bug-eyed conspiracy theories about governments and corporations colluding to
enslave the rest of us.
The Guardian revealed that the US National Security Agency has cracked open
our online lives, that it can rifle through your emails, listen to your calls on Skype,
watching "your ideas form as you type", as a US intelligence officer put it –
apparently in cahoots with the corporate titans of the web.
6
1. Select “I want to paste from the clipboard (Figure 2.1) then press “Next”.
Figure 2.1
2. Paste the text into the space (edit it here if you want).
Figure 2.2
3. Type in a filename for the file (e.g., “Obama1.txt”).
4. Leave “Subcorpus” set to “Add new subcorpus”.
5. Press “Next”. You will be prompted for the name of the subcorpus to add
the file to. Type “News” and then press OK
6. Press “Finalise”.
The file you added should not be displayed in the Project window.(see Figure
2.3).
7
Figure 2.3: Files window after adding a text file
The newly added files are under the caption “Files in corpus but not incorporated
in project”. UAMCT makes a distinction between “incorporated” files, which have
buttons to annotate at all available levels, and “unincorporated” files, which are
in the corpus but not yet opened for annotation.
This distinction is made to make it easy to keep track of those files which you
have started editing, distinct from those you may wish to add later. If you have
100 files in the corpus, but have only annotated five, then you want the five with
annotations to be clearly indicated. This allows for a gradual expansion of your
corpus over time, but let’s you get results at each point.
3 Incorporating Files
To incorporate a file into the project, making it available for annotation, you can
either:
 Click on the “Incorporate All” button to incorporate all unincorporated files,
or
 Click on the “Action” button next to a file and select “Incorporate file” from
the menu. This will incorporate just the single file.
If you do either of the above, you will be presented with a window asking for
some metadata regarding the file or files (See Figure 2.4). This includes:
 Language: which language the text written in? This field is used to
determine which language resources to use for the document. These
resources include lexicons (for concordance searching, calculation of
8


lexical density, etc.), parsers (for automatic segmentation) and taggers.
Currently, only English is really supported, but soon lexical resources for
other languages will be provided.
Encoding: text files are stored in a particular text encoding. You can tell
CorpusTool what encoding your file is in by selecting from this field. The
default option offered by UAMCT is a guess of what it should be, but if the
text does not display properly, you may need to change it. To find out what
encoding the document is in, try right clicking on the document and select
“Open with…” (or the MacOSX equivalent) and open the text with MS
Word, which may help you choose the best encoding. Otherwise, using
‘Open with…’, select a browser, and look for the “Encoding” or “Character
Encoding” menu item, and see which encoding this program gave the text.
Display Font: Choose here the font family and size you want to use to
display your text in the annotation windows. Some fonts will better cope
with non-western writing systems, e.g., some fonts are designed to display
Chinese, etc. However, many modern fonts should display any writing
system.
Figure 2.4: File Metadata Window
After incorporating the file, the Project Window appears as in Figure 2.5.
9
Figure 2.5: The Files Window after incorporating files.
10
Tutorial 3:
Adding a manual annotation layer to your project
The next thing we need to do is to specify what analyses you want in the project.
Let’s start by adding just one layer. A “Layer” is a type of analysis of the text files.
We can add layers for coding clauses, for coding groups, for the register of the
whole text, for appraisal analysis, etc. For this example, we will assume we are
adding a layer for analysing noun phrases (NPs) in terms of both their content
(what they express), and their form (proper, common, pronominal).
1 Change to the Layers pane
Click on the “Layers” button at the top of the window.
2 Click on the “Add Layer” button.
When you click on “Add Layer”, a window will pop up asking several questions.
Use the Next button to move between questions:
1. Layer Name: the name given to the layer. Put “Entity”.
2. Automatic or Manual Annotation: choose ‘Manual’.
3. Scheme: choose “Design Your own”. The other options allow you to use
one of the schemes supplied with UAMCT3, or to use a scheme from
another project you have created.
4. Kind of Segment: here you specify whether you want to assign features
to a text as a whole (e.g., its register or text type) (Whole Document), or
whether you want to assign features to subsegments in the text (e.g.,
clauses). Let’s assume that we are interested in the second, so click on
“Segments within a Document”.
5. Special Layer: This window offers options for special kinds of
annotation. Error annotation layers provide a special slot on the coding
interface for you to provide the correction of the error. RST annotation
provides a special interface for annotating the “rhetorical structure” of
the text. For now, just select “No”.
6. Automatic Segmentation: here you can specify whether the text should
be segmented for you, recognising paragraphs, sentences, or words.
For English texts, automatic recognition of clauses and NPs is also
possible. For this tutorial, select “No”.
7. After following these steps, you will see a final window displaying your
choices, as in Figure 3.1. If any of the settings vary from yours, use the
Back button to go back and change it. Then press “Create Layer” to
return to the main window.
11
Figure 3.1: Last pane of the Add Layer Assistant
Figure 3.2 shows the Layers window with one layer added. The Layer space
provides some information about the layer.
There are two buttons on the Layer control panel:
 Edit Scheme: this button will open a window to allow you to edit the coding
scheme. We will come back to this in the next tutorial.
 Delete Layer: this will delete the layer, and all analyses of text files
performed on this layer. Press this only before you begin coding of the
layer, or if you really want to delete the layer.
 Edit Details: this button is currently disabled, but will in the future allow you
to change the characteristics of the layer (e.g., manual/automatic, autosegment, etc.). Currently, you need to delete the layer and add it again to
change the characteristics.
12
Figure 3.2: The Layers Window with one Layer added
2.1 Return to the Files pane
If you click on the Files button, you will see that the display has changed slightly.
The entry for the “Obama1.txt” file now has a button next to it “Entity”. You can
click on this button to edit this file at this layer. The colour of the annotation
buttons are colour coded to indicate their degree of completeness:
 Light: Not yet coded
 Medium: Partially Coded
 Dark: Coded to a high degree
Don’t open the annotation window just yet, though. First, we need to specify the
coding scheme for the Entity layer. The next tutorial will deal with this process.
13
Tutorial 4:
Editing the Coding Scheme
1 Opening the Scheme Editor
Before annotating files for a given layer, you need to define the annotation
scheme for the layer. The first step here is to open the scheme editor. Change to
the Layers pane, and click on the “Edit Scheme” button for the layer. This tutorial
assumes we are working on the “Entity” layer defined in the previous tutorial.
Figure 4.1: The Entity Scheme before editing
A window like Figure 3.2 will pop up. It shows a small “system network” (a
hierarchy of features), with “entity” as the most basic concept, and a choice
between entity-1 and entity-2.
2 Editing the Scheme
These features have been automatically generated, and we will change them to
more informative names.
 Click on “entity-1”, and a menu will appear with options, as in Figure 4.2.
14
Figure 4.2: Options for Features
For more information on these options, please see the manual. For now, we want
to change “entity-1” to something more plausible. Let’s assume that we want to
code our entities as either ‘human’ or ‘organisation’ (we will ignore NPs that refer
to other sorts of entities).
 Select “Rename Feature”. A window will appear allowing you to edit the
feature name.
 Change the feature text to “human” (in UAMCT, all features are in lower
case, and spaces are not allowed, so if you put capitals, they will be
changed to lower case, and spaces will be substituted for “-“.
 Repeat this process to change “entity-2” to “organisation.
Figure 4.3: Start of the Entity Scheme
After this, you should have the scheme as shown in Figure 4.3. Notice also that
the choice between human and organisation has a name, automatically provided
as “ENTITY-TYPE”. If you want to s rename this choice, you could click on
“ENTITY-TYPE”, and select “Rename System” from the menu which appears.
Rename the system to “SEMANTIC-TYPE”. You can type in lowercase, but
UAMCT will always display system names in upper case.
Lets now add a more delicate distinction between types of organisations. We
want to sub-classify organisations as either company, government or media.
15

Click on ‘organisation’ and select ‘Add System”: a new system (set of
choices) will appear under ‘organisation’, as shown in Figure 4.4.
Figure 4.4: The Entity scheme with subsystem



Rename ‘feature1’ to ‘company’
Rename ‘feature2’ to ‘governement-body’
Click on ‘ORGANISATION-TYPE’ and select ‘Add Feature’, calling this
feature ‘media’
The resulting scheme should look like the scheme in Figure 4.5.
Figure 4.5: Scheme with subsystem renamed
As a next step, we want to code each entity not only by semantic type, but also in
terms of form. We can do this by providing a sub-network in parallel to the
content network.
 Click on ‘entity’ and select ‘Add System’: this will add a system
underneath the original one. The curly bracket indicates that, during
coding, you have to select from both the system above and the system
below. See Figure 4.6.
16
Figure 4.6: Network with parallel system
We now need to edit this new system:
 Click on “ENTITY-TYPE” and select “Rename System”: call the system
“FORM”.
 Click on ‘feature1’ and rename it ‘common’
 Click on ‘feature2’ and rename it ‘proper’
 Click on ‘FORM’ and choose ‘Add Feature’, and call the feature ‘pronoun’.
Figure 4.7: The finished scheme
The resulting network should look like Figure 4.7. Coding Schemes can get quite
complex. They can grow to contain hundreds of choices, with lots of parallel
systems and sub-classifications. However, the smaller the scheme, the quicker
the coding.
This is all we need for now. For more details on editing coding schemes, and
including the schemes in your publications, see the UAMCT3 Manual.
17
Tutorial 5:
Manual Annotation
THE REST OF THIS DOCUMENT STILL NEEDS
TO BE UPDATED FROM VERSION 2.8
1 Annotation Types
CorpusTool currently supports two types of annotation:
1. Code-document: the document as a whole is assigned features. Useful for
defining document language, text-type, register, etc. Also can be used to
code features of the writer (e.g., language proficiency).
2. Code-segments: the user defines segments in the document, and assigns
features to each segment. For instance, clauses, NPs, words, speaker
turns, etc.
Below we will explain how to annotate in both manners.
2 Annotating Code-Document files
Each text file incorporated into your project has a button for each layer of
analysis. If you click on a layer button where the layer was specified as “Code
document”, then a window like that in Figure 4.1 will appear.
18
Figure 5.1: An Annotation window
The code-document window has 4 parts:
1. The Text Frame shows the text file. You can scroll to see the whole text.
2. The ToolBar: giving various actions, such as Save, Close and Help (see
below).
1. The Coding Frame contains three boxes:
a. Selected Features (labelled ‘Assigned’): the features already
assigned to the text. Initially, this will contain one feature, the
leftmost (‘root’) feature of the coding scheme for this layer. As other
features are assigned, they will appear here. You can delete
features by double-clicking on the features in the Selected Features
box. The root feature cannot be deleted, since it applies by default
to all documents.
b. Current Choice: the middle box is a choice which needs to be made
for this document. Double-click on one of the options. That choice
will be moved to the Selected Features box. If there are more
choices in the coding scheme, the next choice will then be
displayed.
c. Gloss Box: If you introduced a gloss for a feature in the scheme
(see Section 3.3 above) then, if you (single-)click on a feature in the
Current Choice box, the gloss will be displayed in this space. This
is useful when you have forgotten what exactly is the coding criteria
for this feature.
19
2. The Comment Frame: In this box, you can type comments about the
current segment, either to remind yourself of some problem, or to
communicate with other people working with the same project. For
instance, one might write: “Is this a material or behavioural clause? Check
with IFG.”
In summary, to code a whole document:
1. Select from the options shown in the Current Choice box until no
options remain.
2. If you make a mistake, double click on features in the Selected
Features box to undo the selection.
3. Close the window and your codings will be saved.
3 Annotating Code-Segment files
When annotating a document at a layer specified as “Code Segments”, the
process is slightly more complex.
Firstly, for the sake of this tutorial, let’s add a new layer to our study.
1. Bring the Project Window to the front
2. Click on the Add Layer button on the right of the screen
3. Call the layer “Participant”
4. Select “Annotate Segments”
5. Select “Do not automatically segment”
6. Select “Create New Scheme”
7. Press the “Finalise” button
Note that this adds a new Layer in the layer space, and also adds a new button
for each incorporated file.
Now, let’s define the scheme for this layer:
1. Click on the Edit button in the space for the Participant Layer.
2. When the scheme window opens, change participant-1 to human and
participant-2 to organisation.
3. Click on “PARTICIPANT-TYPE” and select the option “add feature”.
Type in “other-participant”.
Your network should look like that shown below:
Now, close this window, returning to the Project Window.
Click on the “Participant” button for one of your text files.
This will open an annotation window for the document at this layer. See Figure
4.2.
20
Figure 4.2: Code-segments window
This display differs from that for coding a whole document in that there are more
buttons in the toolbar in the middle. These buttons basically allow you to move
through the segments.
3.1 Making, Moving and Selecting Segments






Make segments by ‘swiping’ text: clicking down at one point in the text
and dragging to the place you want to end the segment, then releasing the
mouse.
Select segment: you can select a segment by clicking on the segment
line which runs under each segment. You can tell which segment the
mouse is over, as the line of the segment is highlighted.
Select next/previous segment: use the < and > buttons in the toolbar to
move around between segments.
Select next/previous incomplete segment: use the << and >> buttons in
the toolbar to move to the next or previous segment which is not totally
coded yet.
Resizing Segments: Select the border of a segment by moving the cursor
over the small border marker (a vertical line) until it goes red to indicate
you are over it. Then click down and drag it where you want to go.
Delete segments: if you create a segment erroneously, you can delete it
by selecting the segment then clicking on the delete button in the toolbar.
Alternatively, hit the Delete key.
21
3.2 Ignoring Segments
Click the Ignore button when a segment is selected, and this segment will not be
used in statistical analyses. Ignore segments are shown in grey in the text
window. The same button can be used to unignore a segment.
4 The “Other Actions” Menu
This menu displays some extra options, depending on the kind of annotation
(whole-document, segments) that you are annotating:
 Edit Scheme: Opens the scheme window for this annotation layer, so that
you can edit the scheme, or add/change the glosses associated with
features.
 Add New Feature: Prompts you to type in the name of a new feature,
which is added to the currently displayed set of choices, and assigned to
this segment.
 Copy Features: Copy the features so far assigned to this segment into
memory.
 Paste Features: Assigns the features previously copied to this segment.
 Resegment Document: Wipes all segmentation of this layer for this
document. Note: this deletes all annotation of the document at this layer.
 Show XML: Displays how the currently open file is stored on disk, in XML
format.
 Show Structure: Switches to an alternative display of the segmentation
interface, which approximates more to the standard structural display of
Functional Linguistics. See figure 4.3.
Figure 4.3: Function Structure Display Mode
22

Show Text Stream: brings up a new window which allows you to view
how choices made change throughout this text. Use the “System to
Graph” menu to select a system to view. Use the “Smoothing” menu to
change the degree of smoothing. With 0 smoothing, each choice is shown
in the sequence it occurs in. Use higher levels of smoothing to better view
how choices are distributed over phases of the text. For instance, in Figure
4.4, the text stream shows that passive clauses occur more strongly at the
beginning of the text, and to lesser degrees later in the text.
Figure 4.4: Text Stream Window
23
Tutorial 4:
Adding a “document” layer to your project
The first thing to do in a new project is to specify what analyses you want in the
project. Let’s start by adding just one layer. For
5 Click on the “Add Layer” button.
A “Layer” is a type of analysis of the text files. We can add layers for coding
clauses, for coding groups, for the register of the whole text, for appraisal
analysis, etc.
Let’s start by adding a Layer for the Register (features which belong to the
document as a whole).
When you click on “Add Layer”, a window will pop up asking several questions,
and use the Next button to move between questions:
 Layer Name: the name given to the layer. Put “Register”.
 Automatic or Manual Annotation: choose ‘Manual’.
 Coding Object: here you specify whether you want to assign
features to a text as a whole (e.g., its register or text type)
(Annotate Document), or whether you want to assign features to
subsegments in the text (e.g., clauses). Let’s assume that we are
interested in the first, and select on “Annotate Document”.
 Coding Scheme: the coding scheme is a description of the
features you want to annotate the text with. You have two options
here:
i. Create New Scheme: In most cases, the user is interested
in making their own coding scheme, representing the
features that they themselves are interested in, organised
in the way they feel they should be. CorpusTool includes
an easy to use interface for creating and modifying these
schemes (see section 3).
ii. Copy Existing Scheme: In some cases, you might reuse a
coding scheme that you developed before, or which was
produced by someone else. CorpusTool ships with a few
schemes predefined, which you could use. One of these is
Peter White’s Appraisal network. Another is based on
Granger’s error annotation scheme.
8. For this tutorial, select “Create new scheme”. Then click on the Finalise
button, and your new layer will be added to the Project Window.
Figure 2.3 shows the Project window with one layer added. The Layer space
provides some information about the layer: it’s name (Register), its type (‘codedocument’), and the name of the scheme associated with the layer
(‘Register.xml’).
24
There are two buttons on the Layer control panel:
 Delete: this will delete the layer, and all analyses of text files performed on
this layer. Press this only before you begin coding of the layer, or if you
really want to delete the layer.
 Edit: this button will open a window to allow you to edit the coding scheme.
We will come back to this in the next section.
Figure 2.3: The Project Window with one Layer added
9.
You can also select ‘Import layer’ from the Project menu to add a layer using a
Systemic Coder study (.cd3 files). See Appendix I for more details.
5.1 Opening an Annotation Window
The remaining buttons on each row each correspond to an annotation layer
defined in your project. Click on the button to open an annotation window for this
file at the specified layer.
Button Colours: The buttons for each layer of a document are colour coded to
indicate their degree of completeness:
 White: totally coded
 Light Blue: Partially Coded
 Dark Blue: Coded to a high degree
Note that these colours are indicative only.
6 Quitting CorpusTool
Note that all changes to a project are automatically saved. If you quit the Project
Management Window (using the X in the top right corner), you quit CorpusTool,
all changes saved.
25
7 Continuing a Project
Once your project is created, the easiest way to open CorpusTool to work on
your project is:
1. Open your project folder on the desktop
2. double-click on the .cptr file (which has a blue globe icon).
CorpusTool will open directly with your Project Window.
UNDO: No undo is currently supported. It will be supported in a later version.
Section 5:
Corpus Search
1 Introduction
The Search Interface is opened by clicking on the Corpus Search button on the
Project Window. Figure 5.1 shows this window.
NOTE: You can also open the Search Window from:
 a Scheme window. Click on a feature and select “Show Examples”.
CorpusTool will open the Search window with all segments marked with
that feature displayed.
 Descriptive or Comparative Feature Statistics: Click on the count field of
any set and the instances which make up the count will be displayed.
26
Figure 5.1: The Corpus Search Window
2 Specifying Search Queries
At the top of the window is a menu-driven widget to define your search query.
For this tutorial, we will use a small project called “Finance”, which can be
downloaded from the CorpusTool website, on the Download page.
1. Simple Feature Search: To search for all segments containing a given
feature, click on the widget at the top left (in Figure 5.1, “person”), and
select a feature from one of your layers. Then press “Show” to see all
instances. Press “Save” to save the search results to a file.
2. More Complex Searching: Click on the small “+” next to the feature
selector to extend your query.
 ‘and’: Allows you to add another feature, and the search will return all
segments containing both the nominated features.
 ‘or: Allows you to add another feature, and the search will return all
segments containing either of the nominated features.
NOTE: and and or cannot be mixed!
 ‘and not’: Allows you to add another feature which should be
excluded, and the search will return all segments containing the first
feature but not the second feature.
 ‘containing segment’: this allows search across layers: it returns all
units tagged with the first feature which contain segments at another
layer tagged with the second feature. For instance, one might search
for ‘finite-clause containing person&subject’, to find all finite clauses
where the segment boundaries totally include a segment at the
participant layer which is coded both person and subject.
27


‘containing string’: this will allow you to find all segments with the
nominated feature which contain a given string. Matching is not case
sensitive.
NOTE: this feature is also used for concordance searching
(searching based on lexical features, wildcard matching, etc. See
below for more details.
‘in segment’: this allows search across layers, specifying that
segments should match only if they are contained within segments at
the second specified layer. For instance, one might search for ‘person
in editorial’ to find segments tagged as person in editorials.
Immediate containment: NOTE: for search queries including
‘containing segment’, ‘containing string’ or ‘in segment’, you can
choose between “immediate” and “anywhere”. The difference is as
follows:
 anywhere: if the containing segment contains the specified segment
or string, it will match.
 immediately: Sometimes users allow units embedded within others
at the same layer, for instance, clauses can be embedded within
other clauses. If you specify 'immediately', then if the contained
segment or string falls within such an embedded unit, it will not
match the units in which the unit is embedded. For instance, with
"[They left because [she was tired]]", a search for: clause
containing immediately 'was' would only match the inner
clause.
3. Combining Complex Searches: One can combine complex searches,
e.g.,
person containing immediately “bush” in finite-clause in
editorial&english
3 Concordance Searching
CorpusTool lets you search for lexical patterns (English only currently for most
features).
3.1 Specifying the Search Query
If you specify “containing string” (see above), you can specify a lexical pattern
instead of a simple string. For example, to find passive clauses, "be%
@participle" will match all segments containing any form of 'be' followed by a
participle verb (-en verb).
Note that the corpus is NOT tagged in terms of part of speech (POS). Rather,
CorpusTool includes a large dictionary of English, and looks up each word in the
dictionary. Because of this, a word will match all POS classes to which it belongs.
For instance “be%” will match all occurrences of “being”, even in the context
where the word is not a verb, e.g., “the being”.
Matching occurs as follows:
Case Insensitive: all searching is case insensitive. Thus ‘Birch’ will match ‘Birch’
and ‘birch’ and “BIRCH’.
The search string consists of a sequence of search tokens separated by a space.
Each search token can be of the following format:
28
1) Literal token: a token not containing *, #, @ or % will match the token itself
only.
2) Wildcard token: if the query token includes an ‘*’, the ‘*’ will match any
number of chars. Thus
 ca*
matches
‘cat’, ‘carburettor’, etc.
 *ed
matches
‘weed’, ‘lived’, etc.
 bro*en matches
‘broken’, ‘Brollerglen’, etc.

3) Match any: a ‘#’ by itself matches any single token.
(The above 3 cases should work for any language where words are divided by
space characters or punctuation)
4) Constraining by class: a wildcard form can be followed with ‘@’ and then a
lexical feature, and the form will match only tokens which, according to the
system’s lexicon, can take that lexical class. E.g.,
 ca*@noun
matches
nouns starting with ‘ca’.
 *ing@mental-projecting matches
mental-projecting verbs ending
with ‘ing’.
An asterisk cannot appear by itself, it must have text either before or after it.
A full list of the lexical features that can be used are in Appendix II, and can
be seen within the tool by selecting “Show Wordclass Network” from the Misc
menu of CorpusTool.
5) General class matching: If no token string is provided before the ‘@’, then the
query form matches all tokens which could represent the specified class. E.g.,
 @noun
matches any noun form
 @verb
matches any verb form
 @adverb
matches any adverb form
 @mental-projecting matches any verb which is classified as mental
 @human-noun
matches any noun classified as a human-noun
6) Inflection matching: ‘%’, at the end of a token indicates that all inflection forms
of the token, which should be a root form, should be matched. Thus,
 break% matches
‘break’, ‘broken’, ‘broke’, ‘breaking’, ‘breaks’
 red%
matches
‘red’, ‘reds’ (noun), ‘redder’ (adj), ‘reddest’
 be%
matches
‘be’, ‘is’, ‘are’, ‘was’, ‘were’, ‘been’, ‘being’
 is%
matches
nothing (only roots can be used)
To constrain the inflection matching to a limited set of inflections, one can add
‘noun’, ‘verb’, ‘adjective’ or ‘pronoun’ after the ‘%’. E.g.,
 red%noun
matches
‘red’, ‘reds’
 red%adjective matches
‘red’, ‘redder’, ‘reddest’
29
Note that wildcards cannot be used within % forms. Nor can the string before
the % be blank.
4 Running a Query
After entering your query, you can hit the “Show” button. If your cursor is in a text
field (Containing String), you can hit the Return Key.
5 Modifying a Query
To change a feature selection, just click on the feature to change it.
To delete any of your search extensions, click on the keyword (“&”, “/”,
“containing”, “in”) and click on “remove”.
6 The Result Space
The white space below the Query space displays the results. Click on a result
and the annotation file containing this segment will be opened at the right place.
The three columns at the left indicate the state of each coding:
 P/- Whether or not the segment is totally coded (P=partial)
 */- whether the segment has a comment associated. Click on the segment
to see the comment.
30
Section 6:
Automating Coding
1. Introduction
The Autocode window allows you to assign features to existing segments using
search patterns. For instance, we can identify passive clauses in English using a
pattern like:
'clause' containing 'be% @participle'
Using the Rule Editor, we define a rule like:
(Note: as with Search, lexical-based search patterns currently work for English)
We can then press the "Show" button, and all instances matching the search
query are shown, with a check-box next to each. We can uncheck any item which
is a false match (not truly a passive). Clicking on "Code Selected" will then assign
the "passive" feature to each of the selected segments.
In this way we can quickly code many of the more common grammatical patterns.
To see a sample of such autocode rules, add a new layer to your project, and
use the scheme included with the system "clauses.xml". This includes rules for
process type (mental, verbal, etc.), voice (active, passive), modality, nonfinite
clauses, etc.
1. Opening Autocoder: Click on the Autocode button on the main window of
CorpusTool.
2. Adding a new rule: To add a new rule, click the "Add" button in the list of
buttons at the top of the Autocode window. A window like the following will
appear:
Select a feature which you want to code automatically. Then specify a
search pattern to use (see section on "Corpus Search" for how to specify a
search query). Then press "Save" to keep this rule in memory.
3. Editing a rule: Click on the Edit button to edit the currently displayed rule.
31
4.
5.
6.
7.
8.
Deleting a rule: Click on the Delete button to delete the currently
displayed rule.
Coding with a Rule: When you have a rule selected, press the Show
button to see all segments which match the search pattern component. A
new toolbar appears with three widgets:
Display All/Agreements/Conflicting/Nonconflicting: selecting from this
list allows you to filter out some of the matches:
 All: shows all of the matches
 Agreements: shows all segments already coded with the specified
feature.
 Conflicting: shows those segments which are already coded with a
feature which conflicts with the feature you are autocoding. For
instance, if autocoding as 'passive', this would show all segments
already coded as 'active'.
 Nonconficting: shows all segments which are neither agreements nor
conflicting.
Select All/None: selecting one of these options will select/deselect the
check boxes next to each segment.
Code Selected: Clicking on this button will automatically code all
displayed segments which are selected.
Hints
 For some grammatical phenomena, you can provide a pair of rules like:
 Select passive if contains 'be% @participle'
 Select active if clauses and not passive
 Use the first rule to code passives and then use the second rule to put
everything else as active.
 Provide one rule such as the passive rule above. Code these. Then edit the
rule, inserting a # between the search terms, e.g.,
 Select passive if contains 'be% # @participle'
 This will find some instances where 'not' or an adverb falls between the verbs.
32
Section 7:
Corpus Statistics
1 Introduction
The Corpus Statistics pane allows various statistics to be derived from your
tagged corpus. Press the “Statistics” tab on the main window’s toolbar to see the
Statistics pane (as in Figure 7.1).
Figure 7.1: The Statistics Pane
You can use this interface to perform two kinds of studies on your corpus:
1. General Text Statistics: offers general statistics of the corpus, such as
total number of segments, number of words per segment, lexical density in
the corpus, pronominal usage, etc.
2. Feature usage: you specify a feature in a layer (most typically, the root
feature of the layer), and the program describes the usage of features in
the corpus at that layer (counts, mean, and standard deviation).
33
These studies can be done for a single dataset (descriptive statistics), two
datasets
(comparative statistics), or showing results for each document
individually:
1. Describe a dataset: offers descriptions of your corpus, or a specified
subcorpus.
2. Compare two datasets: provides a comparison of two subsets of your
corpus (e.g., english vs. spanish). When Feature is selected, the two sets
are contrasted in terms of the occurrence of presence of the features in
the codings at the layer specified. Levels of significance of the differences
between the sets are displayed, both in terms of Students T-test and ChiSquared (see below).
3. Compare Multiple Files: provides details of each file in your corpus, one
column per file.
2 A Contrastive Feature Study
Figure 7.2 shows a sample Comparative study done using the “Finance” project.
Note very little of the text has been annotated, so the results are for small
numbers only. We would need to tag a thousand or more participants from
a range of editorials and fpn (front page news) articles before we could
start to trust the results. This preliminary study shows two significant
results (more reference to people rather than organisations at a 98% level
of significance; and significant differences in the types of organisations
discussed), but the numbers are too low to trust.
Figure 7.2: A Contrastive Stats Study
34
3 Performing a Study
To perform one of the studies outlined above:
1. Choose one of the options from the “Type of Study” menu: ‘describe a
dataset’, ‘compare two datasets’ or ‘compare multiple files’.
2. Choose from the “Aspect of Interest” menu: choose either ‘Feature
Coding” or “General Text Statistics”.
3. Specify the unit that you are interested in (see section 5, part 2: Specifying
Search Queries). This should be the unit which you wish to explore
differences in. It could be the root feature in a network (as in the case in
Figure 7.2), or a more delicate one.
4. If you are selected “Compare two datasets”, then enter a feature in the Set 1
space and another in the Set 2 space. This should be a unit which
CONTAINS the unit of interest. In this case, we specify units of the Register
layer, fpn and editorial. Since these features apply to whole texts, they do
contain the segments with feature “participants”.
5. Press Show.
4 Interpreting the Results: Feature-based Studies
Only systems which are relevant are shown. For instance, if we had specified the
unit of interest as “person” above, then the study would involve only those
segments with feature “person”. For this reason, the results for this system are
not shown, as “person” would score 100%, and the other features in that system
0%.
Counts and Percentages: The results for each feature are shown with both raw
counts (how often that feature occurred in the dataset) and also as a percent.
The percent shows the proportion of segments which have this feature. Note that
the percentages in a system (a given set of choices) always adds up to 100%, so
really what it is measuring the propensity to select this particular feature as
opposed to the other features in the same system.
Statistical Significance: when a comparative study is done, it is possible to
measure whether the differences between the two datasets is statistically
significant (does it represent a real difference or is it possibly due to randomness
in the data).
CorpusTool uses two measures of statistical significance, and presents them
both in the results:
 T-Statistic: T-Stats are the numbers on which the level of significance of
your result can be derived. The bigger it is, the higher the level of significance,
but this also depends on how much data you have. In some more scientific
papers, you might be requested to provide T-Stats, but it is quite rare in
linguistics.
 Chi Squared: in recent years, particularly in linguistics, chi squared statistics
are becoming the preferred means of testing significance. CorpusTool
provides the Chi Squared statistics for each comparison, and the level of
significance that corresponds to this.
At the end of each entry there will be between 0 and 3 "+" signs. These indicate
how statistically significant is the difference of this features mean from that of the
mean of the other set:
(none)
Not significantly different.
+
Significant at the 90% level (10% chance of error).
35
++
Significant at the 95% level (5% chance of error).
+++
Significant at the 98% level (2% chance of error).
The level of significance is important to establish how repeatable your results are.
Results without significance may be accidents, and if we repeat the study with
other texts, the result may be different. If results are highly significant they are
likely to be repeatable if we apply the analysis to a totally different set of texts.
To understand this, a single + means that of any 10 such results, you can expect
one to be a false result (90% significance, or 10% chance of error).
5 Presenting Results as a Network
When performing a feature-based study, you can now view the results in a
system network, instead of in table format. See Figure 7.3. After a study is
presented in table form, a new menu is presented, labelled “View As”. Select
“Network” to switch to Network view.
This way of displaying statistics has been copied from a similar feature in
SysFan1. I thank Canzhong Wu, the author of SysFan, for allowing me to use this
feature.
1
Available from http://minerva.ling.mq.edu.au/units/tools/index.htm
36
Figure 7.3: Network View of Statistics
6 Saving Statistics
Each Statistics window offers a “Save” button which allows you to save the
results to file, in HTML format, tabbed delimited, or plain text.
Results saved in HTML can be opened in MS Word, and then cut/pasted into
your publications.
Results saved tab-delimited can be opened in MS Excel (on Windows, right-click
on the .txt file and specify Open with… Excel.) These files may also be useful for
programs such as SPSS.
37
Section 8: Keywords
1 Keywords
The top words in any frequency list for English will be words such as “the”, “of”
and “a”. A more informative listing works out how important each word is for a
particular corpus, when compared with a more general corpus.
For instance, the keywords from a corpus split over three fields are shown below.
The words are ordered in terms of their “specialness” for this corpus (relative
frequency in this corpus when compared to the relative frequency in the general
corpus). A value of 100 indicates the word appears 100 times more in this corpus
than in other corpora.
NOTE: for this to work, one needs to select only a sub-corpus. If you select the
whole corpus, then nothing will happen.
Military
troops
100.0
weapons
100.0
engine
100.0
mountains 100.0
smoke
90.0
gulf
85.0
enemy
85.0
aircraft
80.0
force
70.0
civilians 70.0
civilian
70.0
guys
65.0
military
62.47
squadron
60.0
suicide
55.0
tanks
55.0
soldier
55.0
jungle
55.0
altitude
55.0
strikes
55.0
trees
55.0
lieutenant 55.0
withdrawal 55.0
missile
55.0
bomber
50.0
invasion
50.0
combat
50.0
rounds
50.0
missions
45.0
Economics
economy
121.38
companies 116.52
stock
100.0
tax
100.0
cuts
85.0
profits
80.0
investment 75.0
billion
75.0
returns
70.0
sales
70.0
earnings
65.0
investors 65.0
jobs 65.0
package
65.0
assets
65.0
prices
60.0
bill
60.0
corporate 60.0
stocks
58.26
markets
55.0
budget
55.0
finance
50.0
volatility 50.0
reforms
45.0
commercial 40.0
temporary 40.0
cent
37.87
analysts
32.04
growth
32.04
Crime
crime
142.85
detective 50.0
police
49.16
disappearance
40.0
criminal
39.86
court
34.88
justice
30.23
driver
30.23
boy
29.06
victims
18.6
family
17.56
child
13.95
car
12.81
lived
11.96
officers
11.96
legal
11.51
children
10.57
kids
9.3
mercy
9.3
investigators
9.3
woman
9.01
murder
8.52
boys
8.52
age
7.77
victim
6.64
street
6.27
body
6.22
incident
5.98
2 Phrases
Rather than looking at single-words, n-gram analysis looks for sequences of
words which are common in the corpus. For instance, a list of the frequent 338
grams (sequence of 3 words) that occur in a small corpus of introductions to
academic papers are shown below:
in terms of
a set of
in this paper
the performance of
of the two
be able to
a number of
the design of
which can be
the problem of
12
11
10
7
7
7
7
7
7
6
ad hoc networks
we believe that
of this paper
terms of a
in section 4
some of the
in order to
large number of
that can be
ad hoc network
6
6
6
5
5
5
5
5
5
5
According to Biber (e.g., Biber and Barbieri 2007), as the corpus grows to a
reasonable size (millions of words), the kinds of phrases that raise to the top
don’t contain lexical content as such (e.g., ‘ad hoc networks’). Rather, they are
phrases which are used to frame such meanings. We see here: “in terms of”, “a
set of”, etc.
While keywords tell us which words we should teach in a text, n-grams can tell us
which phrasings are usefully taught. For instance, assuming we were teaching
students how to write introduction sections to academic papers, we collect a
corpus of such texts and produce the key n-grams for various lengths. From such
a corpus, we can pick up frequent phrases such as “this paper reports on” or “this
paper/article is organized as follows”.
3 Key Features
‘Key Features’ does what ‘keywords’ does, except that, rather than looking at the
words in the text, it looks at the features assigned to segments. The software
thus shows which features are special to the focus corpus, as compared with the
reference corpus.
A key value of ‘2.0’ indicates that the feature is twice as common on the focus
corpus as in the reference corpus.
39
Section 9: Text Styling
4 Text Styling
It is sometimes useful to view the coding of a text visually. CorpusTool allows you
to view one of the text files of your project, specifying that particular segments
(on whichever layer) should be showed in bold, italic, underline, larger font or
coloured. See Figure 8.1 for the text style view of a file within the “Finance”
project.
Figure 8.1: The Text Styler Window
5 Opening the Text Styler
From the Project window (the main window), click on the filename of one of your
files. Note, this only works for files incorporated in your project. Also, your project
needs to have at least one layer defined.
6 Styling the Text
You can assign colour and/or font effects (bold, italic, underline) to all text tagged
with a given feature, or feature combination. This allows the patterns of selection
throughout the text to be visible.
E.g., use bold/italic/underline for appraisal categories, and colour coding for
clause type to see how appraisal is distributed in respect to clause types.
40
7 Saving Styled Text
You can save styled text to an HTML file. To include styled text in an MS Word
document, open the HTML file in MS Word, and from there cut/pasted into your
own document.
41
Section 10: The Menubar
1.
Merge Projects
Up until Version 2.8.4, this function did not work correctly.
Before Merging: To ensure cleaner merger, delete any layers in either project
which have no annotation associated with them. In all cases where the same files
exist in both projects (e.g., schemes, corpus files and annotation files), the file in
the current project will be preserved. Files will moved from the other project
where they do not exist in the current project.



Select “Merge projects” from the Project menu.
You will be asked to select another project to merge with the current one.
Results will be saved in a new folder with the same name as the currrent
project, plus "-merged" added.
42
Appendix I:
Importing Systemic Coder Studies
1 How to Import Coder Studies
The analysis files in Systemic Coder can be imported into CorpusTool. To do so,
follow the following instructions:
If you have a single file to import:
1. Ensure that the coding scheme is saved as an external file (master
scheme). To do this, open the file in Coder, and select “Scheme Storing…”
from the Options menu. Select “Save to Master” and specify a location to
save the scheme.
2. Ensure the codings are saved as .cd3 not .cd2: if the file on disk has a
.cd2 extension, you need to open the file and select “Save Codings As”
from the File menu. The program will offer to save it as a .cd3 file.
3. Now, make a new folder and place within it the scheme file and the
codings file.
4. Open CorpusTool and create a new project.
5. Select “Import Layer” from the Project Menu.
6. You will be asked to specify the folder created in (3) above.
7. The .cd3 file will be split into the raw text (to be put into your Corpus
folder) and the analyses (placed in the Analyses folder). The next window
asks in which subcorpus folder to place your text file.
8. The analysis scheme is imported as a new layer. The next window asks
for the name of the layer.
9. In Coder, the only way not to code a bit of text was to ignore it. In
CorpusTool, one selects only the bits of text one wants to code. You may
thus want the ignored segments in your Coder study to disappear. The
next window allows you to do this.
10. Press Finalise, and you have a new Layer added, and your cd3 file is
imported.
If you have a set of files, all annotated with the same scheme:
1. Place all the Coder files in a folder.
2. Make sure ALL the files are in .cd3 format, not .cd2.
3. Follow step (1) for a single file, for at least ONE of the files (e.g., make
sure there is a .scheme file in the folder)
4. Proceed from step (4) for the single file case above.
43
If you have one or more files, where the same text(s) have been coded with
different networks (in a sense you have done multi-layered annotation using
Coder):
1. For each set of files annotated with the same scheme, create a folder and
place the coder files and the scheme file for that analysis. (ensure the files
are in .cd3 format).
2. Ensure all files which are analyses of the same file have the same file
name, e.g., if you have Text1-CLAUSE.cd3 analysed for clauses, and
Text1-GROUP.cd3 analysed for groups, rename both files to Text1.cd3.
(CorpusTool can only tell two files are analyses of the same text by having
the same filename).
3. Open a new project and use the Import Layer option as described above
for one of the folders.
4. Repeat (3) for the other folders.
Some problems may arise:
 CorpusTool says it cannot read one or more of your .cd3 files: it
may contain characters which are outside of ASCII text. CorpusTool
should handle this, but currently cannot. Send me your files and I
will import it for you.
 If you have any other problem importing cd3 files, send them to me
(make a zip of the folder) and I will look at it (this is good for me, to
see the kinds of problems people are having, so I can fix them).
44
Appendix II:
Lexical Features for Concordance Searching
1 Nouns
COMMON- singular-noun
INFLECTION plural-noun
reg-noun
regd-noun
glreg-noun
COMMON-NOUNinv-noun
REG-PATTERNS
fixedsingular
fixedplural
apostr-s
thing-noun
temporal-noun
common-noun
event-noun
institution-noun
report-noun
COMMONsubstance-noun
TYPE
place-noun
human-noun
disease-noun
quality-noun
group-noun
noun
NOUNTYPE
countable-noun
COMMONmass-noun
NOUN-TYPE2
collective-noun
given-name
PERSONNAMEperson-name
surname
TYPE
title
determiner-required
PLACENAME-TYPE determiner-not-required
place-name
proper-noun
PROPERTYPE
country-name
PLACEplanet-name
NAME-TYPE2
city-or-state-name
medication-name
day
TIMEtime-name
month
NAME-TYPE
time
organisation-name
religion-name
45
 Verbs
infinitive
thirdpersonsingular
LEXVERBpast-verb
INFLECTION
participle
gerund
aux-verb
inflecting-verb
AUXVERB-TYPE
do-aux
have-aux
AUX-VERB- reduced-aux
TYPE2
nonreduced-aux
intransitive-verb
INFLECTINGVERB-TYPE
doing-verb
DOING- monotransitive-verb...
TYPE
ditransitive-verb...
link-verb
LEXlex-verb
VERB-TYPE
mental-verb...
PROJECTINGprojecting-verb
verbal-verb...
VERB-TYPE
appearance-verb
relating-verb
INFLECTING-VERB- reg-verb
REG-PATTERNS
regd-verb
infinitive-be
verb
VERBTYPE
1ps-be
2ps-be
3ps-be
BE_AUXpres-plur-be
INFLECTION
past-sing-be
past-plur-be
be-aux
participle-be
gerund-be
reduced-be
BEAUX-TYPE nonreduced-be
modal-verb
reduced-modal
MODALVERB-TYPE nonreduced-modal
Subclasses of mental verb
cognition-verb
COGNITIVE- cognitive-action-verb
VERB-TYPE cognitive-state-verb
MENTALVERB-TYPE reaction-verb
please-type-verb
Subclasses of verbal verb
mental-verb
addressee-oriented-verb
verbal-verb
VERBALVERB-TYPE
telling-verb
ADDRESSEEORIENTED-VERB-TYPE2 asking-verb
proposing-verb
ADDRESSEEnot-addressee-oriented-verb
ORIENTED-VERB-TYPE
stating-saying-verb
STATING-SAYING- stating-verb
VERB-TYPE
saying-verb
saying-required
SAYINGVERB-TYPE saying-not-required
46
 Adjectives
er-est-adjective
absolute-adjective
ER-EST-ADJcomparative-adjective
INFLECTION
superlative-adjective
ER-EST-ADJECTIVE- reg-adj
REG-PATTERNS
regd-adj
ADJECTIVETYPE
noninflecting-adjective
adjective
NONINFLECTING- comparable-adjective
ADJECTIVE-TYPE noncomparable-adjective
ADJECTIVE- nationality-adjective
SEMTYPE
other-adjective
 Adverbs
manner-adverb
temporal-descriptive
descriptive-adverb
DESCRIPTIVE- connective-adverb
ADVERB-TYPE2 modal-adverb
location-adverb
other-descriptive-adverb
more-less
ADVINTENSIFIERadverb
most-least
TYPE intensifier TYPE
degree-intensifier
jussative
interrogative-adverb
ago-adverb
47
 Pronouns
nominative-pronoun
accusative-pronoun
PRONOUNgenitive-pronoun
CASE
genitive2-pronoun
reflexive-pronoun
1p-pronoun
personal-pronoun
PRONOUN- 2p-pronoun
PERSON
3p-pronoun
pronoun
wh-personal-pronoun
3Phuman-pronoun
PRONOUN-TYPE
neuter-pronoun
PRONOUN- singular-pronoun
NUMBER
plural-pronoun
NOUNWC
spatial-pronoun
temporal-pronoun
LOCATIONthing-pronoun
PRONOUN-TYPE
one-pronoun
nonpersonal-pronoun
that-pronoun
wh-pronoun
LOCATIONPRONOUN-TYPE2 nonwh-pronoun
 Number
cardinal
NUMnumber
ordinal
TYPE
percentage
 Conjunction
conjunction
CONJTYPE
and-conjunction
COORDcoordinating-conjunction
or-conjunction
TYPE
but-conjunction
pre-or-infix-conjunction
subordinating-conjunction
NORMAL-SUBORDINATING- infix-only-conjunction
CONJUNCTION-TYPE
if-conjunction
gerund-conjoiner
 Prepositions
agent-preposition
to-preposition
preposition
PREP- of-preposition
TYPE as-preposition
by-preposition
other-preposition
48
 Determiners
DETTYPE2
positive-strict-determiner
STRICTstrict-determiner
negative-strict-determiner
DETERMINER-TYPE
wh-strict-determiner
nonstrict-determiner
determiner
quantifying-determiner
NONSTRICTDETERMINER-TYPE demonstrative-determiner
singular-determiner
QUANTIFYINGDETERMINER-NUMBER plural-determiner
 Punctuation
period
SENTENCEsentence-final-punctuation
exclamation-mark
FINAL-TYPE
question-mark
comma
punctuation
CONJUNCTIVE- semicolon
PUNCTUATION- conjunctive-punctuation
TYPE
hyphen
TYPE
colon
open-bracket
bracketing-punctuation
BRACKETING- close-bracket
single-quote
TYPE
double-quote
open-quote
DOUBLEQUOTE-TYPE close-quote
genitive-s
49